Generative AI for Facial Expressions in 3D Game Characters: A Retrieval-Augmented Approach

This paper examines a Retrieval-Augmented Generation (RAG) approach for generating facial expressions in 3D game characters using artificial intelligence. By integrating large language models within a RAG-based architecture, we developed a proof-of-concept system that animates expressions based on Facial Action Coding System (FACs) action units. Testing demonstrates the potential of RAG-driven animations to create immersive, adaptive experiences with contextually appropriate expressions that enhance perceived emotional responsiveness in Non-Playable Characters, highlighting RAG’s promise for dynamic character interactions and AI-driven personalization in game development.